This is a first draft comments are welcome!

Political Representation and Citizen Involvement. The Social Policy Responsiveness to Different Participants in Europe 1 Yvette Peters Bergen University Abstract Political participation has been argued to be important in order to ensure political representation. Moreover, unequal participation has been feared to result in unequal representation. While some studies have examined this link, this paper aims to disentangle in what way participation can result in policy responsiveness through 1) different mechanisms (promissory and anticipatory representation), and 2) different causes (the representatives aim to be re-elected and the communication of preferences to representatives). This is done by analyzing the relationship between preferences of different groups of participants (activists; voters; alternative participants; non-participants) and policy output. Furthermore I consider the individual characteristics of these participants in order to find whether there are systematic differences. This is a first draft comments are welcome! 1 This paper is prepared for presentation at the ECPR General Conference, September 3-6, 2014, Glasgow 1

INTRODUCTION Political participation is important for democracy. One crucial reason for its importance is that the communication of preferences enables the possibility for the representation of those preferences. At the same time, it has been shown that political involvement among citizens is not equal. Most notable income, education, gender and age help determine who participates (e.g. Armingeon & Schädel, 2014; Marien, Hooghe, & Quintelier, 2010; Verba, Schlozman, & Brady, 1995). This inequality in participation has led to the assumption that representation is also unequal, and has caused concerns about democracy (Lijphart, 1997; Verba, 2003; Verba et al., 1995). Research in the field of representation has illustrated the existence of such differential representation, showing that the poor (e.g. Bartels, 2008; Gilens, 2012; Rosset, Giger, & Bernauer, 2013), women (Bühlmann & Schädel, 2012; Enns & Wlezien, 2011; Reynolds, 1999), and minorities (Griffin & Newman, 2007; Reynolds, 2013) appear to be systematically underrepresented. Furthermore, while it has often been argued that participation is necessary for representation, only some studies have explicitly connected these concepts, showing that voters and the more engaged are better represented (Adams & Ezrow, 2009; Griffin & Newman, 2005; Martin & Claibourn, 2013). I aim to extend this line of research connecting participation and representation by looking at differential policy responsiveness in Europe, including four types of citizens: activists (those who vote and undertake at least one more political act), voters (those who only vote), alternative participants (those who do not vote but do undertake another political action), and non-participants (those who do not participate at all). Moreover, I distinguish between anticipatory and promissory representation (Mansbridge, 2003), where anticipatory (or dynamic; Stimson, MacKuen, & Erikson, 1995) representation implies responsiveness between elections to ensure re-election; and promissory representation is the responsiveness on the basis of electoral promises (or manifestos). My analyses suggest that voters are first of all responded to through promissory representation: the government that was selected or composed on the basis of their votes follows their manifesto-promised direction (left-right) in terms of social spending. Moreover, citizens who only vote as a form of participation are also responded to in an anticipatory way, provided that the group of only-voters is substantial in size. People that are more active, however, appear to be even more influential in determining social policies, also when their 2

groups is relatively small. Non-participants appear under-represented in the area of social policies, however. POLITICAL PARTICIPATION AND REPRESENTATION Political participation is a core aspect in an ideal democratic process people within a democratic unity should have the opportunity to discuss their ideas and make them known to each other, they should have the possibility to vote on decisions, and they should be able to gather enough information and insights in order to make an informed decision (Dahl, 1998). Even though there is no agreement about how much participation is needed for democracy with ideas ranging from minimizing participation to the election of a government (Schumpeter, [1943] 1954), to expanding the range of participation and influence of citizens widely (see e.g. Pateman, 1970; Warren, 1993) few would challenge participation as a crucial aspect of democracy. Indeed, participation is often considered the elixir of life for democracy (Van Deth, 2014; 350). Political participation is necessary, too, in order for governments and policies to reflect to some extent what people want. Active political participation can help ensure more accurate preference-policy congruence. Griffin and Newman (2005) explicitly studied the link between participation and representation, and found that voters preferences indeed better predict Senators roll-call behavior in the United States. Similarly, Martin and Claibourn (2013) find that policy responsiveness is increased in districts with higher levels of turnout. Adams and Ezrow (2009), on the other hand, look at the group of citizens that regularly discuss politics and try to persuade others of their viewpoints. They argue that this group will be more visible and is better able to communicate its preferences, and find that these opinion leaders influence party programs in Western Europe, while other citizens do not. This study aims to contribute to this line of research. By distinguishing between different groups of participants, and by taking into account government ideology, I aim to tap into two assumptions concerning the relation between participation and representation. First, it has been assumed that voters are also more active in communicating their preferences to representatives between elections and are responded to partly because of that (Adams & Ezrow, 2009; Griffin & Newman, 2005). By examining the responsiveness to people that only vote and people that vote and undertake alternative activities as well, it can be shown to what extent it 3

matters whether people are more or less active, and whether it matters what type of participant someone is. Second, it has been argued that voters are better represented partly because they choose likeminded representatives. Moreover they argue that voters will be able to reward or punish them in the next election, and that they communicate their preferences as a continuous process between elections (Griffin & Newman, 2005). These arguments suggest two representative mechanisms; anticipatory (responsiveness between elections) and promissory responsiveness on the basis of electoral promises) representation (Mansbridge, 2003). By including government ideology to indicate representation through casted votes, as well as people s preferences to indicate more continuous representation, I disentangle these two mechanisms. Inequality in participation and representation One important concern related to the democratic process, linking participation and representation, is that previous studies have shown that there are systematic biases in who participates. For example, income, education, gender and age have been found relatively good predictors of (different types of) political participation (e.g. Armingeon & Schädel, 2014; Marien et al., 2010; Stolle & Hooghe, 2005; Verba et al., 1995). If participation ensures more accurate representation, such biases are likely reflected in responsiveness, resulting in unequal representation. This, in turn, directly affects the realization of the democratic ideals of political equality (see Lijphart, 1997; Verba, 2003; Verba et al., 1995). Examining differences in congruence and responsiveness in various contexts, research has shown that the preferences the less affluent appear to be taken less or not at all into account (Bartels, 2008; Enns & Wlezien, 2011; Gilens, 2005; Rosset et al., 2013), for example. Moreover, Griffin and Newman (2007) more explicitly examine the link between unequal participation and representation in the United States, and find that increased participation among Latino s will also increase their representation. Thus, unequal participation may result in unequal representation. This dilemma is somewhat further complicated by the use of alternative channels of participation, i.e. not voting. Voting offers citizens a relatively easy way to voice their preferences, where each citizen has the opportunity to cast a vote within a set time period at a nearby location, where information and debates are offered in a relative centralized way, and where votes are counted and weighed relatively equally. Extra-representative forms of participation (Teorell, Torcal, & Montero, 4

2007) such as petition signing or demonstrating, on the other hand, are often much less centralized also in terms of information, and can be more ad-hoc (Stolle, Hooghe, & Micheletti, 2005). Compared to voting, these types of participation offer citizen a less equal opportunity to voice preferences, and voices are not necessarily weighed equally if they are considered by representatives or policy-makers. So, while expressions of preferences are necessary, both in terms of electing a representative body and signaling between elections, the seemingly inherent inequality in political involvement also poses a threat to this democratic process. Therefore, I will also briefly consider which characteristics the different types of participants that I examine here have, and whether they are different from each other. HYPOTHESES There are several reasons why the preferences of active citizens are likely to be better represented than those of non-participants (see also Adams & Ezrow, 2009; Griffin & Newman, 2005). First of all, voters whether they undertake other activities or not will tend to elect parties or candidates whose ideology best correspond to their preferences, and governments are composed on the basis of such elections. Ideally, then, government policies would largely follow the preferences of the majority of the voters, on the basis of the promises made before the election. Voters would therefore be better represented than non-voters, who did not express a preference for who should be in office. Assuming that voters are able to elect a candidate or party that also matches their general preferences (left-right) for policies, I expect that the leftright position of the government affects that government s policies. More specifically, I expect that more left-wing governments are associated with higher social spending, than more rightwing governments. This selection-based expectation captures the mechanism of the aforementioned promissory representation: the idea that representation takes place on the basis of pre-election promises. This leads to my first hypothesis: H1: Promissory representation hypothesis: Left-wing governments have higher levels of social spending than right-wing governments. Second, in order to ensure also anticipatory representation, representatives should also respond to people s preferences more dynamically between elections. Here, the idea is that citizens can 5

sanction a representative by voting him or her out of office. Assuming that representatives or parties are indeed motivated by re-election, they would attempt to represent voters not (just) on the basis of their promises from the previous election, but (also) on the basis of voters preferences between elections (Mansbridge, 2003). This would mean that representatives are specifically interested in responding to the preferences of voters (Adams & Ezrow, 2009; Griffin & Newman, 2005) in a more dynamic way, than following the voters ideological cues as would be the case in promissory representation. This is captured in the following hypothesis: H2: Anticipatory representation hypothesis: Social spending as an expression of social policies is more responsive to voters preferences than to non-voters preferences. Lastly, it can be argued that representatives need information about people s preferences beyond regular elections. In this theoretical argument, it is assumed that representatives aim to represent preferences, though are not necessarily motivated by the aim of re-election. Citizens need to communicate their preferences to policy-makers in order for dynamic responsiveness to be possible at all. Indeed, it has been argued that voters and opinion leaders are also more active between election, ensuring that their preferences are communicated and represented (Adams & Ezrow, 2009; Griffin & Newman, 2005). This argument implies that people who are more active citizens (alternative participants) should be better responded to than people that are less active, or than people that only vote. More specifically, people that undertake an alternative form of participation are expected to be better represented than people that do not: H3: Information representation hypothesis: Social spending as an expression of social policies is more responsive to alternative participants preferences than to non-alternative participants preferences. These hypotheses are not mutually exclusive, and should not be seen in isolation from each other. While the three theoretical arguments have largely been the basis for the expectation that voters or opinion-leaders are better represented (Adams & Ezrow, 2009; Griffin & Newman, 2005; Martin & Claibourn, 2013), I aim to test these hypotheses in a somewhat more nuanced way. First, H1 and H2 both suggest that voters are better represented (with or without additional political activities), but the two point at two different mechanisms (promissory and 6

anticipatory). Moreover, it is possible that representatives respond according to both mechanisms. They can follow a general direction (left-right) that is given to them by voters in t1, while also adapting policies to preferences of voters between elections in order to ensure reelection at t2, within the margin of the general direction given earlier. Second, H2 and H3 are both concerned with the anticipatory or dynamic mechanism of representation, but make a distinction in who the representative focuses on. On the one hand, if representative are interested in re-election they would be specifically interested in voters and less so in people that do not vote (even if they are active in other ways). On the other hand, the representative needs some kind of input in order to be able to take into account citizens preferences voting alone does not seem to be enough from this perspective. Combing these expectations about who the representative focuses on, especially active citizens those who vote and undertake (an) other action(s) would be influential in determining policies. This is the group that representatives would aim to satisfy if they are motivated by re-election, and the group that helps the representative by communicating their preferences. Voting, then, may appear to be a necessary condition for representation, both in anticipatory and promissory mechanisms of representation. Moreover, since this group of citizens has the possibility to sanction the representative out of office, voting may also be a sufficient condition for representation, as long as the representative has some way of getting information on voters preferences. Alternative forms of participation, however, may be necessary for accurate representation simply because of the communication of preferences, but may not be a sufficient condition because these citizens do not necessarily also vote (and would thus not have the possibility to sanction). MEASUREMENT AND METHOD OF ANALYSIS Policy responsiveness is in effect the relationship between people s preferences and some policy output (see also Soroka & Wlezien, 2010). I focus here on social policies in terms of social spending and economic inequality. To measure social spending, which is the dependent variable in this study, I use data provided by the Organisation for Economic Co-operation and Development (OECD), and is expressed as the amount of social spending as a percentage of GDP. Concerning people s preferences, I employ the first five waves of the European Social Survey (ESS; 2002-2010). The surveys ask people to what extent they agree with the following statement: The government should take measures to reduce differences in income levels. The (recoded) 7

answer categories for the respondents are disagree strongly, disagree, neither agree nor disagree, agree, and agree strongly. While this question does not explicitly ask whether people want more or less spending, it can nonetheless be interpreted as such. When respondents answer that they agree with the statement, they indicate that the government should make some kind of effort to reduce income differences, and this can practically be expressed by more social spending by the government. Policy responsiveness is then the expression of the relationship between social spending and people s preferences for more or less redistribution, over time (2002-2010, every two years) and across countries (Austria, Belgium, Switzerland, Czech Republic, Germany, Denmark, Estonia, Spain, Finland, France, United Kingdom, Greece, Hungary, Ireland, Israel, Italy, Luxembourg, Netherlands, Norway, Poland, Portugal, Sweden, Slovakia, Slovenia). The effective number of cases in my sample is 106. In order to compare different groups of participants to each other, I calculate the average preference on reducing differences in income for each group, per country and year. These groups are composed on the basis of the type of political activity individuals have undertaken, something the ESS also provides data on. I distinguish between two types of participation: voting (yes/no) and alternative participation (yes/no), the latter implying that an individual signed a petition, boycotted a product, and/or participated in a demonstration. Combining these two types of participation result in four groups of individuals: people that vote and undertake some other activity, people that only vote, people that only undertake some other activity, and people that do neither. Figure 1 illustrates this typology. Figure 1: Four groups of participants Alternative form No alternative form Vote Activist Voter No vote Alternative participant Non-participant 8

The size of these groups within each country and year are not equal in size. It is therefore important to account for this in order to be able to compare relative responsiveness I include the size of each group as an interaction term. This way, I eliminate the possibility that one group is better responded to simply because it is a larger group, by comparing responsiveness levels at similar group size levels. Furthermore, it would be interesting to see whether a more sizable group is indeed also more important in terms of predicting levels of social spending: when a group is composed of more than 50% of its citizens, it also constitutes a majority. One might expect that a majority is more likely to get what it wants than a minority. Moreover, since the dependent variable is social spending as a percentage of GDP I take into account the growth in GDP as a control variable. Social spending as measured here, is dependent on the level of GDP when it increases from one year to another the percentage social spending will automatically go down if the amount of social spending remains the same. To control for such dependence, I include the annual growth in percentage GDP, using data from the World Bank. Lastly, in order to tap into the idea that general representation takes place on the basis of the promises that parties made before the election, I include a measure that indicates the ideology of the government. For this, I use data from the Comparative Manifesto Project, where party positions are coded on several issues over time, and for a wide range of countries. They also provide a left-right positioning including several of these issues, where higher negative scores indicate more left-wing positions and higher positive scores indicate more right-wing positions. I calculated the weighed left-right position for the government, relative to size of the parties that were included, in order to indicate government ideology. The indicator empirically ranges from 37.81 (most left) to 32.71 (most right). Based on this type of ideology, I thus expect that a negative relation between ideology and social spending, indicating that left-wing parties are associated with higher social spending than right-wing parties. The promissory representation hypothesis, however, is based on the assumption that government ideology is indeed a reflection of voters ideology. When government ideology would have the expected effect on social spending, it implies that parties promises from before the elections ring through policies. It does not necessarily mean, however, that this is also a reflection of voters ideology. It is possible, for example, that people vote on the basis of little or incorrect information, or that people vote on a certain party simply because s/he wants to express some form of discontent and with that, the casted vote might not accurately reflect 9

the voters ideology. Therefore, I test to what extent voters and non-voters ideology are related to government ideology. The ESS includes a question asking respondents to place themselves on a political left-right scale, ranging from 0 to 10. Using this measure, the average placement can be calculated for both voters and non-voters, per country and year. While the values of the left-right scales are not identical, they run in the same direction: the higher the score, the more right-wing; and the lower the score, the more left-wing. Table 1 shows the results of a simple model including government ideology as the dependent variable, and the ideology of (non-) voters as independent variables. It shows that the political position of voters is positively related to government ideology, while that of non-voters is negatively related: voters indeed appear to determine government s general policy position on a broad left-right scale. Table 1: the relation between government and (non-) voters ideology Ideology of voters 8.452** (4.168) Ideology of non-voters 14.588** (5.048) Intercept 26.776 (19.211) R 2 0.059 N countries 24 N 105 Notes: Results from a Prais-Winsten regression analysis, with panel corrected standard errors reported in parentheses. The model Israel 2008, because this case is an outlier with have too high leverage on the overall results (although including the case does not change the results in a substantial way). ** p<0.05, two-tailed The Anticipatory representation and information representation hypotheses rely on the idea that policies are in dynamic accordance with people s preferences, depending on what type of participant they are. Representation here implies a positive relation between preferences and social spending: the more people support the idea that the government should take measures to reduce inequality, the more social spending should increase. However, since the indicator for preferences includes support for both more and for less effort to reduce inequality, I model the relationship between preferences and policy output in a quadratic fashion. This allows for responsiveness to mean a negative relationship between spending and preferences when people want less government effort, while at the same time it means a positive relationship when people 10

want more government effort. Moreover, the relation should be stronger when people that indicate that the strongly (dis) agree, than when people simply indicate (dis) agree. Responsiveness would thus follow a U-shaped relation, where the turning point of this relation is around the middle score of the preference index, implying a preference that supports neither more nor less spending. Lastly, since it takes some time to make and implement policies, it is unlikely that social spending responds instantly to preferences. Therefore, and in line with most research investigating responsiveness, all dependent variables are lagged with one year. These considerations lead to the following generalized model specification: Social spending = β +β ( ) +β ( ) +β! ("#$) +β % ( "#$) +β ' ( "#$) +β ( ()*+ #,-.) +β / (012 3+h) (1) The structure of the data, with cases being countries at a specific time points, requires a crosssectional time-series modelling approach. The assumptions for ordinary least squares have been tested for, and the results of these tests also form the basis for the model choice. Models showed persistent serial correlation, panel heteroscedasticity and for some models there appeared cases that were outliers which also had too high leverage on the overall results, but there were no signs of a unit root. I therefore employ a Prais-Winsten regression analyses with panel corrected standard errors to correct for the serial correlation and heteroscedasticity, and excluded the cases that could not be predicted by the model and had too high leverage which cases these are, is reported with the respective models. As a second step, I analyze to what extent the different groups are composed of different groups, and whether there are biases on the basis of socio-economic characteristics. This is done to examine the claims of unequal representation on the basis of unequal participation. In this analysis, I include several individual characteristics, which have been identified as (possible) predictors of participation: age, gender, income, education, and status of employment. This way, it can be determined whether some characteristics make it more likely that someone is a 11

member of a certain participant group (activist, voter, alternative participant and nonparticipant). For this, I again employ the first five waves of the ESS, which also has information about individual characteristics. Since these analyses have individuals as cases, I analyze the data with a multilevel logit maximum likelihood estimation process, where the country is considered the higher level and time is included to control for temporal trends. The variables included are: age and age squared (age in years), foreign born (0=born in country of residence, 1=not born in country of residence), female (0=male, 1=female), education (1=less than lower secondary education, 2=lower secondary education completed, 3=upper secondary education completed, 4=post-secondary non-tertiary education completed, 5=tertiary education completed), income (1-10 index, where 1 is the lowest category, and 10 is the highest category), employed (0=not employed, 1=employed (the phrasing in the ESS refers to paid work during the last 7 days )), and time trend (1=2002, 2=2004, 3=2006, 4=2008, 5=2010). RESULTS Table 2 shows the results of the parameter estimation of equation (1), where each model indicates the representation of one of the four participant groups. First, it shows that the coefficient of government ideology is negative and statistically significant for all four models. This implies that more left-wing governments are associated with higher levels of social spending, while more right-wing governments tend to be accompanied by lower spending levels. This finding is in line with the idea that representation takes place on the basis of voters electing like-minded parties or candidates: their ideology is reflected in the ideology of the government, and the government tends to execute the promises (as put down in their manifestos) they made before the election. The coefficient of government ideology ranges between 0.049 and 0.073, which is relatively substantial considering that the theoretical range of the left-right index is 100 to 100, and the empirical range is between 37.8 and 32.7. For example, a shift from a right-wing government scoring 15.0 to a left-wing government scoring 15.0 would imply an increase of between 1.47% and 2.19% in social spending. This finding supports the promissory representation hypothesis in the sense that governments appear to implement policies in accordance with the ideology that formed the basis for people s votes. Furthermore, the models in Table 2 show that the growth in GDP is negatively related to social spending. As expected, the level of social spending as a percentage of GDP depends on the level of percentage growth in GDP, simply because growth in GDP does not automatically 12

Table 2: The participatory effects of popular preference and ideology on social spending Activist Voter Alternative participant Nonparticipant Model 1 Model 2 Model 4 Model 4 Group 16.692 75.824 25.190 40.005*** (22.771) (62.503) (18.453) (13.249) Group^2 2.111 8.510 2.747 5.122*** (3.073) (8.031) (2.356) (1.864) Size group 444.236** 506.638** 438.076 503.493* (218.248) (239.377) (1,222.614) (272.474) Group*size 238.774** 243.888** 109.320 234.626* (118.137) (122.351) (640.864) (133.211) Group^2*size 32.925** 28.710* 2.333 28.509* (16.047) (15.527) (84.017) (16.276) GDP growth 0.497*** 0.541*** 0.718*** 0.480*** (0.101) (0.100) (0.139) (0.115) Government ideology 0.049** 0.073*** 0.065*** 0.053** (0.025) (0.025) (0.025) (0.023) Intercept 12.000 138.471 80.298** 105.538*** (42.260) (121.066) (36.188) (23.916) R 2 0.834 0.795 0.764 0.794 N countries 24 24 24 24 N 105 106 104 104 Notes: Models report the results of Prais-Winsten regression analyses, with panel corrected standard errors reported in parentheses. All independent variables are lagged with one year. Model 1 excludes Israel 2010, Model 3 excludes Estonia 2008 and Greece 2010, and Model 4 excludes Greece 2004 and 2008; since these cases could not be predicted by the model and have too high leverage on the overall results. * p<0.1; ** p<0.05; *** p<0.01, two-tailed mean that the amount of spending also grows proportionally. An increase in GDP would reflect a lower level of social spending as a percentage of GDP, even if the actual amount spent could be the same or even somewhat higher than before. The results in Table 2 concerning the more anticipatory type of representation are illustrated in Figure 2. It shows the marginal effects of preferences on social spending (responsiveness) along the aggregate indication of the extent to which people (dis)agree that the government should take measures to reduce inequalities, for each type of participant and various group sizes. Thus, the y-axis can be interpreted to indicate responsiveness, and the x- axis as the preference of the group for less (1 and 2) or more (4 and 5) spending. The upper left graph shows the responsiveness for participants people who vote and undertake an alternative action. It shows that when this group s preferences indicate that they want less spending, the relationship between preferences and social spending is negative: when 13

Figure 2: Marginal effects of preferences and group size on social spending, for each group of participants Marginal effect of preference for redistribution on social spending -100-50 0 50 100 * * * * Strongly 1 2 3 4 Strongly 5 disagree agree Activist preference Marginal effect of preference for redistribution on social spending -100-50 0 50 * * * * * * Strongly 1 2 3 4 Strongly 5 disagree agree Voter only preference Group size=10% Group size=25% Group size=10% Group size=25% Group size=51% Group size=75% Group size=51% Group size=75% Marginal effect of preference for redistribution on social spending 0 20 40 60 80 100 * * * * * * * Marginal effect of preference for redistribution on social spending -50 0 50 100 Strongly 1 2 3 4 Strongly 5 disagree agree Alternative participant preference Group size=10% Group size=25% Strongly 1 2 3 4 Strongly 5 disagree agree Non-participant preference Group size=10% Group size=25% Group size=51% Group size=75% Group size=51% Group size=75% Notes: based on models 1 to 4; * indicates p<0.1, two-tailed this groups wants less spending, their preferences seem to influence social spending to go down. On the other hand, when participants want more spending, their preference indicates an increase in spending. This pattern follows the expectation of how the relationship between preferences and social spending would develop if responsiveness exists. While the change in the direction of the relationship is not at the middle category (neither more nor less spending), it 14

roughly follows the U-shaped relation between preferences and policy output as described above. Furthermore, it shows that this pattern holds, even when the group is relatively small. When the participant group is composed of at least 25% of the population, it appears that their preferences are reflected in policies, suggesting at least some level of responsiveness. This description of the relationship between preferences and policy output is only partly similar for the group that only votes, shown in the upper right corner of Figure 2. Considering that responsiveness remains negative also when the group expresses mild support for more spending, it appears that there is a bias towards groups with preferences for less spending among voters. Moreover, the group of only voters seems to be responded to only when their group composes at least about 51% of the population. Smaller groups of only voters do not see their preferences much reflected in policies. The group of alternative participants, on the other hand, hardly seems to be represented at all. The lower left quadrant in Figure 2 illustrates that especially for the groups that want less spending (1 and 2), there is no relationship between preferences and spending. The preferences of alternative participants are only moderately responded to when these preferences indicate that they want an increase in spending but also here, the relationship is not clear-cut, as this does not seem to be the case when that preference is very strong (a score of 5 on the x-axis). Overall, it does not seem to matter whether the size of the group of alternative participants is very large or small; this group s preferences are generally not well reflected in policy output. Lastly, the lower right corner of Figure 2 illustrates the results for the non-participants. It shows that the relationship between preferences and policy output is never statistically significant for none of the preference categories, or for any of the group sizes. Nonparticipants preferences do thus not appear to be reflected in policy output. These results seem to indicate that the act of voting is important in order to be represented. People part of the group that only votes appear to be largely responded to in terms of social policy output, provided that this group is large enough. Moreover, preferences of the group in which people vote and undertake an alternative form of participation help determine levels of social spending, also when the group is relatively small. On the other hand, responsiveness to alternative participants seems low, and that of non-participants non-existing. These findings support the anticipatory representation hypothesis in that voters appear better represented than 15

non-voters. It underlines the idea that voters are important to representatives, suggesting that they aim to represent voters better since they can vote them out or back in office. Furthermore, the results provide some support for the information representation hypothesis. Responsiveness seems more prominent for activists than for voters only, even though responsiveness to alternative participants only is marginal. It suggests that participants can make it easier for representatives to be responsive when they signal their preferences. That alone does not, however, seem to be a sufficient condition for representation. It simply aids better representation. Combining these two hypotheses and the results, it appears that voting is both necessary and sufficient as a condition for representation. It suggests that representatives consider voters important and target them specifically in creating and implementing policies. When voting is combined with alternative participation, however, groups can further improve their representation. They help representatives to respond to their preferences by communicating their preferences. With that, alternative participation is neither necessary nor sufficient as a condition for responsiveness, although can help improve groups preferences to be reflected in policies, especially when the members of the group also vote. Communication between citizens and representatives is thus not necessary for representation, although it helps to improve it. Lastly, I consider the possible socio-economic inequalities that may be incorporated in the differential responsiveness to types of participants. Table 3 shows the results of the multilevel analysis, including several individual characteristics to predict whether a respondent is a member of a certain participant group (activist, voter, alternative participant and nonparticipant). The first results column in Table 3 illustrates that women, the higher educated, the rich and employed are present in the activist group. Moreover, it shows that age is initially positively related to being an activist where it becomes more likely that someone is a participant the older (s)he gets and that this relationship reverses at a later age (around 55 years). The second column shows that people that only vote are generally more likely to be older (age is almost linearly related to being a voter, with the slope only decreasing slowly at a higher age), male, lower educated and employed. Income does not appear to have an effect on being a voter only. Alternative participants (third column) are more likely to be younger, higher educated, and to have a lower income neither gender nor employment seems to have an effect. Lastly, non-participants are generally younger, lower educated and have a lower income. 16

Table 3: Individual characteristics and participation Activist Voter Alternative participant Nonparticipant Age 0.086*** 0.062*** 0.004 0.041*** (0.002) (0.002) (0.005) (0.002) Age^2 0.001*** 0.000*** 0.000*** 0.000*** (0.000) (0.000) (0.000) (0.000) Female 0.136*** 0.096*** 0.033 0.019 (0.013) (0.011) (0.026) (0.016) Education 0.082*** 0.024*** 0.012*** 0.170*** (0.004) (0.003) (0.004) (0.007) Income 0.085*** 0.002 0.060*** 0.119*** (0.003) (0.003) (0.006) (0.004) Employed 0.188*** 0.140*** 0.045 0.014 (0.016) (0.014) (0.031) (0.019) Time trend 0.017*** 0.016*** 0.029*** 0.051*** (0.005) (0.004) (0.009) (0.006) Constant 4.020*** 2.130*** 2.262*** 0.349*** (0.144) (0.120) (0.167) (0.123) ICC 0.115 (0.030) 0.083 (0.022) 0.120 (0.031) 0.081 (0.022) N countries 24 24 24 24 N 147754 147754 147754 147754 Notes: Models report the results of logit multilevel maximum likelihood analyses, with standard errors reported in parentheses. *<0.10; ** p<0.05; *** p<0.01 These results suggest that especially people who are younger, have a lower income and are not employed are less likely to be part of one of the groups that are responded to activists and voters. While men and women, as well as the higher and lower educated are more likely to be members of one of these two groups, 2 the younger, poorer and unemployed citizens seem to be under-represented within the represented groups while they are over-represented in the less represented groups (alternative participants and non-participants). These people are less likely to be a member of a participant group whose preferences are related to policy outcome. DISCUSSION This study has examined the differential responsiveness to different types of participants. Since participation has often been assumed to form the basis of representation, structural inequalities 2 This does not mean that men and women, or the higher and lower educated are exactly equally represented this also depends on the size of the participant group. However, the results suggest that neither group necessarily is necessarily less likely to be a member of a group that is better represented. 17

in who participates has caused some concerns for the health and functioning of democracy. Indeed, previous research has shown that voters and opinion-leaders are more likely to be taken into account in the formation of policies, than people who are less politically active. My study has aimed to contribute to this field of research by 1) specifying and testing the mechanisms by which participants are argued to be better represented, and 2) by more explicitly connecting representation and unequal participation. The results suggest, by and large, that voting is important in order to be represented. First, voters ideology is largely reflected in the ideology of the government and the government s positions help predict the level of social spending as an expression of social policies. The generalized policy positions that governmental parties set in their manifestos is also reflected in the policies they shape: left-wing governments tend to spend more related to social policies than right-wing governments. Indeed, the findings point at promissory representation the idea that representatives do what they promised to do before the elections. Second, voters seem to be important for representatives and political parties, also encouraging anticipatory representation. Since voters are the ones that are able to sanction representatives, it is relevant for them to consider preferences of voters. The findings suggest that policies are related to what voters want, supporting this argument. The relation between preferences and spending appears strongest for the activist and the voter only groups, although the latter only when the group is relatively large. Furthermore, representation seems to be helped by undertaking also other forms of participation communicating preferences to the representative facilitates responsiveness. This, however, does not appear to be either a necessary not sufficient condition. Since voters appear to be the crucial group, their representation is aided by voicing preferences between elections and besides the act of voting itself. This study thus highlights the importance of political participation especially voting for the well-functioning of representation. However, as noted earlier, the structural differences in who participates and who does not have caused concerns about the political equality of citizens (Lijphart, 1997; Verba, 2003; Verba et al., 1995). If the people that participate are of specific groups within society, these groups would also have a smaller chance to be represented and are thus not equally considered. Even though representation does not appear equal to each group, the results show that the different groups of participants indeed under- or over-represent people on the basis of age, gender, education, income and employment. Both men 18

and women, as well as higher and lower educated people, are disproportionally present in one of two better represented groups, and does not lead to the suggestion that any of these people are better or worse represented. Since this could also depend on the size of the participant group, future research should investigate these differences further. However, especially the younger, poorer and unemployed citizens are less likely to be part of one of the groups that is better responded to. Here, it is likely that these citizens are also under-represented. This finding further underlines the findings from previous research that some groups in society, most notably the poor, are under-represented (e.g. Bartels, 2008; Gilens, 2012; Rosset et al., 2013). Moreover, other research has found that a lack of responsiveness either in terms of party positions or party ideology also reduces the chances that people turn out to vote (Dinas, Trechsel, & Vassil, forthcoming; Lefkofridi, Giger, & Gallego, 2013). This means that a lack of responsiveness (so some groups) would set in motion a vicious cycle: people would be less likely to participate, and with that they reduce again their chances to be represented. Lijphart (1997) and Verba (2003) strongly emphasize the importance of participation in particular voting for the realization of an increasingly equal democratic system. From a democratic point of view, citizens should be considered political equals, and in addition these equal citizens should be able to set the policies that govern them, at least to some extent: while not all people can get what they want all the time, there cannot be any systematic biases in this influence. It is thus important to help and motivate the inactive to become active, at least through the means of voting. One other way to promote equal representation is to institutionalize a form of descriptive representation, where the parliament would mirror society in personal characteristics. Such practices have been argued to, and sometimes shown to improve the quality of representation (Banducci, Donovan, & Karp, 2004; Mansbridge, 1999; Murray, 2014; Reynolds, 2013), and might help to equalize representation. 19

LITERATURE Adams, J., & Ezrow, L. (2009). Who Do Europeans Parties Represent? How Western European Parties Represent the Policy Preferences of Opinion Leaders. Journal of Politics, 71(1), 206-223. Armingeon, K., & Schädel, L. (2014). Social Inequality in Political Participation: The Dark Sides of Individualisation. West European Politics, 1-27. doi: 10.1080/01402382.2014.929341 Banducci, S. A., Donovan, T., & Karp, J. A. (2004). Minority Representation, Empowerment, and Participation. Journal of Politics, 66(2), 534-556. doi: 10.1111/j.1468-2508.2004.00163.x Bartels, L. M. (2008). Unequal Democracy. The Political Economy of the Gilded Age. Princeton: Princeton University Press. Bühlmann, M., & Schädel, L. (2012). REPRESENTATION MATTERS: THE IMPACT OF DESCRIPTIVE WOMEN'S REPRESENTATION ON THE POLITICAL INVOLVEMENT OF WOMEN. Representation, 48(1), 101-114. doi: 10.1080/00344893.2012.653246 Dahl, R. A. (1998). On Democracy. New Haven: Yale University Press. Dinas, E., Trechsel, A. H., & Vassil, K. (forthcoming). A look into the mirror preferences, representation and electoral participation. Electoral Studies(0). doi: http://dx.doi.org/10.1016/j.electstud.2014.04.011 Enns, P. K., & Wlezien, C. (Eds.). (2011). Who Gets Represented?. New York: Russell Sage Foundation. Gilens, M. (2005). Inequality and Democratic Responsiveness. Public Opinion Quarterly, 69(5), 778-796. Gilens, M. (2012). Affluence and Influence. Economic Inequality and Democratic Responsiveness. Princeton: Princeton University Press. Griffin, J. D., & Newman, B. (2005). Are Voters Better Represented? Journal of Politics, 67(4), 1206-1227. Griffin, J. D., & Newman, B. (2007). The Unequal Representation of Latinos and Whites. Journal of Politics, 69(4), 1032-1046. doi: 10.1111/j.1468-2508.2007.00605.x Lefkofridi, Z., Giger, N., & Gallego, A. (2013). Electoral Participation in Pursuit of Policy Representation: Ideological Congruence and Voter Turnout. Journal of Elections, Public Opinion and Parties, 24(3), 291-311. doi: 10.1080/17457289.2013.846347 Lijphart, A. (1997). Unequal Participation: Democracy's Unresolved Dilemma. American Political Science Review, 91(1), 1-14. Mansbridge, J. (1999). Should Blacks Represent Blacks and Women Represent Women? A Contingent "Yes". Journal of Politics, 61(3), 628-657. Mansbridge, J. (2003). Rethinking Representation. American Political Science Review, 97(4), 515-528. Marien, S., Hooghe, M., & Quintelier, E. (2010). Inequalities in Non-institutionalised Forms of Political Participation: A Multi-level Analysis of 25 Countries. Political Studies, 58, 187-213. 20

Martin, P. S., & Claibourn, M. P. (2013). Citizen Participation and Congressional Responsiveness: New Evidence that Participation Matters. Legislative Studies Quarterly, XXXVIII(1), 59-81. Murray, R. (2014). Quotas for Men: Reframing Gender Quotas as a Means of Improving Representation for All. American Political Science Review, 108(03), 520-532. doi: doi:10.1017/s0003055414000239 Pateman, C. (1970). Participation and Democratic Theory. Cambridge: Cambridge University Press. Reynolds, A. (1999). Women in the Legislatures and Executives of the World: Knocking at the Highest Glass Ceiling. World Politics, 51(04), 547-572. doi: doi:10.1017/s0043887100009254 Reynolds, A. (2013). Representation and Rights: The Impact of LGBT Legislators in Comparative Perspective. American Political Science Review, 107(2), 259-274. Rosset, J., Giger, N., & Bernauer, J. (2013). More Money, Fewer Problems? Cross-Level Effects of Economic Deprivation on Political Representation. West European Politics. Schumpeter, J. A. ([1943] 1954). Capitalism, Socialism, and Democracy (4th ed.). London: Allen & Unwin, 4th edition. Soroka, S. N., & Wlezien, C. (2010). Degrees of Democracy. Politics, Public Opinion, and Policy. Cambridge: Cambridge University Press. Stimson, J. A., MacKuen, M. B., & Erikson, R. S. (1995). Dynamic Representation. American Political Science Review, 89(3), 543-565. Stolle, D., & Hooghe, M. (2005). Shifting Inequalities? Patterns of Exclusion and Inclusion in Emerging Forms of Political Participation. Paper presented at the 101st Annual Meeting of the American Political Scienc Association, Washington D. C. Stolle, D., Hooghe, M., & Micheletti, M. (2005). Politics in the Supermarket: Political Consumerism as a form of Political Participation. International Political Science Review, 26(3), 245-269. Teorell, J., Torcal, M., & Montero, J. R. (2007). Political Participation: Mapping the Terrain. In J. W. Deth, J. R. Montero & A. Westholm (Eds.), Citizenship and Involvement in European Democracies; A Comparative Analysis (pp. 334-357). London: Routledge. Van Deth, J. (2014). A Conceptual Map of Political Participation. Acta Politica, 49(3), 349-367. Verba, S. (2003). Would the Dream of Political Equality Turn Out to be a Nightmare? Perspectives on Politics, 1, 663-679. Verba, S., Schlozman, K. L., & Brady, H. E. (1995). Voice and Equality. Civic Voluntarism in American Politics. Cambridge, Massachusetts: Harvard University Press. Warren, M. E. (1993). Can Participatory Democracy Produce Better Selves? Psychological Dimensions of Habermas's Discursive Model of Democracy. Political Psychology, 14(2), 209-234. doi: 10.2307/3791409 21